The multimodality of irony has attracted attention from empirical scholars, whose primary concern has been with the role of multimodal markers of irony for identification and interpretation of irony in natural communication. However, scarce systematic research has been done on the construction, mechanism, and representation patterns of verbal-pictorial irony. This paper, drawing on theories such as conceptual metaphor, metonymy, and blending in cognitive linguistics, constructs an online blending model by integrating metonymy, metaphor, and mini-narrative. This model, plausibly compliant with the cognitive reality in processing image-text irony, provides great explanatory power. Besides the model, it constructs the conflict representation system of multimodal irony. And examples are employed to illustrate the varying categories of representation. The multimodal irony framework established serves in promoting the study of online cognitive operation for multimodal irony.